Modulated Vortex Lattice in High Fields and Gap Nodes

Modulated Vortex Lattice in High Fields and Gap Nodes

O

o

<

o
o

Oh

X3
O

>

oo
m

o

o

I

o
o

X

Ryusuke Ikeda^ and Hiroto Adachi^
^Department of Physics, Kyoto University, Kyoto 606-8502, Japan
^Department of Physics, Okayama University, Okayama 700-8530, Japan
(Dated: February 2, 2008)

The mean field vortex phase diagram of a quasi two dimensional superconductor with a nodal
d-wave pairing and with strong Pauli spin depairing is studied in the parallel field case in order
to examine an effect of gap nodes on the stability of a Fulde-Ferrell-Larkin-Ovchinnikov (FFLO)-
like vortex lattice. We find through a heuristic argument and a model calculation with a four-fold
anisotropic Fermi surface that the FFLO-like state is relatively suppressed as the field approaches
a nodal direction. When taking account of a couple of experimental results altogether, the present
result strongly suggests that the pairing symmetry of CeCoIns should be of d^y-type.

PACS numbers: 74.20.Rp, 74.25. Dw, 74.70.Tx

In a recent paper (denoted as I hereafter), we
have examined the vortex phase diagram of quasi two-
dimensional (Q2D) type II superconductors with strong
Pauli paramagnetic (spin) depairing by focusing on H || c
case with a field H perpendicular to the superconducting
layers. In contrast to earlier works taking account

of both the orbital and spin depairing effects of the mag-
netic field in clean limit, the orbital depairing has been
incorporated fully and nonperturbatively there 0, and
two new results opposite to those suggested previously
0,0] were found. First of all, the mean field (MF) transi-
tion at the Hc2 (T)-line changes from the familiar second-
order one to a first-order (MF-FOT) 0, H, 111 one at a
higher temperature T* than a region in which a Fulde-
Ferrell-Larkin-Ovchinnikov (FFLO)-hke modulated
vortex lattice may appear. This feature is consistent with
data of CeCoIns in H || c H H i, 0. Second of all, a
second order transition curve ffpFLO (T) between such a
FFLO-like and ordinary vortex lattices remarkably de-
creases upon cooling. Interestingly, these two results are
also consistent with more recent data of CeCoIns, sug-
gesting a structural transition to a FFLO state, in H _L c
[3, IjOlllll H^ . A recent ultrasound measurement
also shows that the suggested FFLO state is, as we have
argued in I, a kind of vortex lattice. However, it should
be further examined theoretically whether this qualita-
tive agreement with the data in H _L c is justified or
not.

In this paper, results of an application of analysis in I
to a model for the H _L c case are reported. By includ-
ing the contributions, neglected in previous works
from the nonCaussian (|A(r)|'* and |A(r)|^) terms of the
Ginzburg-Landau (GL) free energy to the spatial gradi-
ent parallel to H, where A(r) is the pair-field, we find
that the relative position between T* and ifpFLO-line is
qualitatively the same as in H || c case as far as a
spin depairing strength realistic in bulk superconductors
is used, and that, at least close to i/pFLOj the LO state

with periodic nodal planes perpendicular to H of
|A| is more stable than the FF state 0,0 composed of
a phase-modulation with keeping |A| fixed.

A special attention is paid in this paper to a notice-
able in-plane angular dependence of the FFLO curve
Hvv T,n(T) found in specific heat and magnetization
data of CeCoIng: The observed FFLO curve in H \\
[110] lies at higher temperatures than that in 11 [100].
This i/pFLO-anisotropy is much more remakable [9| than
that of Hc2 (T) and may give a decisive information on the
four-fold anisotropy of the gap function. As far as an in-
plane Fermi velocity anisotropy is negligible, it is heuris-
tically predicted through the following simple argument
how a gap anisotropy results in a -ffpFLO-a-nisotropy:
Near the gap nodes where the superconducting gap Ak
is small, the coherence length £k — Tivp/Ak defined lo-
cally in the k-space is longer [13]. An orbital limiting
field -fforb(O) is inversely proportional to the square of
an averaged coherence length in the plane perpendicular
to H and hence, is minimal when H is directed along
the four-fold symmetric gap nodes (or minima). Since a
higher Hoih will lead to a relatively stronger effect of spin
depairing, the FFLO curve and T*, induced by the spin
depairing, are expected to lie at higher temperatures as
H is located along a gap maximum. If comparing the
expected iJppLo-anisotropy with the observations 9, 1^
in CeCoIns, we inevitably reach the conclusion that, in
agreement not with the original argument 4] favoring a
(i2.2_j^2-pairing just like in high-Tc cuprates but with a re-
cent report on low H specific heat data Q , a node (or a
minimum) of the gap function of CeCoIns is located along
the [100]-direction, implying a d^jy-pairing state. Below,
we will show how this conclusion is reinforced through a
microscopic derivation of -ffpFLo(2^) taking account of a
possible in-plane four-fold anisotropy of the Fermi surface
(FS). The present result might require a serious change in
the picture on the pairing mechanism of CeCoIns based
upon similarities on the normal state properties, includ-
ing the presence of antiferromagnetic fluctuation, to the
high Tc cuprates .

2

First, let us sketch the outhne of MF analysis for
H II c. Throughout this paper, we assume H = Hx and
the d-wave gap function — \/2cos(20) or V2sin(20),
where (p is the azimuthal angle in the a-h plane. Within
the lowest {N = 0) Landau level (LL), the GL free energy
density in the MF approximation takes the form

ao(Q)(|Ag)p) + ^(|AS)|^)

^(|aS¥>

C0+C2g'+C4Q^ (1)

The essential part of the MF analysis in I is to derive
the coefficients, ag, V4, Vg, C2, and C4 by starting from
the weak-coupling BCS model with a Zeeman (Pauli-
paramagnetic) term. Here, N{Q) is an averaged density
of states (DOS) at the Fermi level, and ( ) is the spa-
tial average on y and z. A(r) was expanded in terms
of the LLs as A(r) = Ag^'(2/, z)wQ(a;), and the

higher LLs were neglected above. For the LO (FF) state,
uq{x) takes the form cofi{Qx) (exp(iQa;)). A Q2D FS
with a circular form in y-z plane was assumed, although
an in-plane anisotropy will be conveniently included as 0-
dependences of the Fermi velocity and DOS (See eq.(5)
below). For an example, a^iQ) is, after performing k-
integrals and introducing a parameter integral, expressed
by

iV(0)ao(Q) = I^u*q{x)(^j^^^2t:T

where ( )x denotes the spatial average on x, pqH is the
Zeeman energy, and A^(0)|(7| is the dimensionless pairing
interaction strength. The function g^'^\p^—idx) has the
form

so that the resulting Hyvlo{T) is a second order tran-
sition line. If the effective strength of spin depairing
^o-fforb/(27rfcB'7co) IS of Order unity or larger, a phase
diagram derived numerically in this manner includes a
^^FFLO (7')-line decreasing upon cooling, where /ig-ff is
the Zeeman energy, and is the orbital limiting field
in 2D limit. In Ref.l where the V4 and Vg-contributions
to C2 were neglected, the LO and FF states had the
same -ffpFLO-line, while we find that the instability of
the straight vortex lattice into the LO vortex state [j, 1^
occurs at a slightly higher tempeature than that into the
FF state 0,0 ■ Hence, at least close to iJFFLo(r), the LO
state becomes the ground state in -ffFFLO < H < Hc2-
Further, we find that the Ve-contribution to C2 is quan-
titatively negligible, while the ifpFLO-line is pushed by
the corresponding V4-contribution down to a lower tem-
perature region in which iJc2 and the vortex state just
below it are described by the iV = 1 LL. Thus, at least
within the weak-coupling BCS model, a FFLO state in
H II c rarely occurs because such a iV = 1 LL vortex
lattice has no FFLO-like modulation X]. We guess that
a slight specific heat anomaly |3] in CeCoIns in H || c at
low enough temperatures may be rather due to a transi-
tion between straight vortex lattices in = and A^ = 1
LLs. A detailed study of this transition into a A^ = 1 LL
state will be reported elsewhere.

Now, let us turn to the H _L c case. Although, in
principle, the above analysis can be extended to a Q2D
system with a cylindrical FS under H perpendicular to
the cylindrical axis, we have chosen to work in an ellip-
tic FS elongated along z[\\ c)-axis and with the disper-
sion relation = h"^ J2j=x,y,z^i^^'] / C^"^) under H || x
in order to make numerical calculations more tractable,
where Jx = Jy = 7""'^^^, and 7z = 7 with 7 > 1 and
a constant m. We expect the case with a moderately
large 7-value to qualitatively describe essential features
in the realistic Q2D case. By isotropize the k vector
as kj = jjkpfj, where f — (cos(/)sin0, sin0sin0, cos0) is
the unit vector in the spherical coordinate, the veloc-

ity V on FS is written as Vj = 7^ "^^FT^j- A Jacobian

g'^^^Pj—idx) — N{0)exp{—p'^Vp/Ar'j^)cos{—ipvFdxl?) \/ jsin^O + ^~'^cos^9 accompanying the angular integral

along FS is exactly cancelled by the angular dependence

where rn is the magnetic length, and vp the Fermi ve-
locity. The extension, oat, of oq to the A^-th LL is given
by multiplying eq.(3) by £N{p'^v'j,/2rjj), if keeping just
terms diagonal with respect to the LLs, where Cisi{x)
the A^-th Laguerre polynomial. The coefficients Vi{Q)
and V(,{Q) are derived in a similar manner to above. The
coefficients C2 and C4 arise from the Q-dependences of oq,
Vi, and Vq.

The onset T* of MF-FOT at Hc2 is determined by
V4(0) = irrespective of the details of higher order non-
Gaussian terms of the GL free energy, while -ffFFLoC?^)
is defined by C2 = under the condition C4 > 0. We
have verified that the latter condition is always satis-
fied throughout the computations in the present work.

of DOS N(e) = N{Q)vf/ ^Y.jv'^r Again, the in-plane
(four-fold) anisotropy of FS will be first neglected. Then,
the GL free energy within A^ = LL takes the form of
eq.(l), and the function g'^^\p, —'^dx) appearing in ao((3)
(see eq.(3)) is replaced in the present case by

(0)/

idx) =

X coa{-ipVxdx), (4)

where Vy^ — 777

77? ^vl-

The parameter fj is
insensitive to the uniaxial anisotropy 7 but dependent
on T and needs to be determined by maximizing Hc2(T).
By focusing on the low T region, we find that fj takes

3

0.5 0.6

FIG. 1: H-T mean field phase diagram obtained using 7 =
3 and witli no in-plane FS anisotropy. The transition or
crossover positions in H || gap- maximum (|| gap-node or mini-
mum) are expressed by the solid curves and filled circle (chain
curves and open circle). The dotted curve and open triangle
denote, respectively, the MF-FOT line and the position at
which the two solid curves ajv(O) = in iV = and 1 merge
with each other.

a value between 0.4 and 0.5 depending on the relative
angle between H and a nearest nodal direction. Using
this parameter, the anisotropy in spatial variations of
A(r) within the y-z plane is given by 7/77. Except for
modifications indicated above, the corresponding quartic
and 6th order terms of the GL free energy are derived
by closely following the analysis in I. We choose a\\ —

/^o-ffori~^^(^)/^B7cO as a measure of the spin depairing
strength in H _L c, where hI^^^^\q) is the orbital limiting
field in the isotropic case.

In Fig.l, the resulting phase diagram is shown to illus-
trate how the i?FFLo('7^)-position depend upon the rela-
tive angle between H and the nodal directions. Thin solid
(chain) curves are defined by aAr(O) = 0, and the Hc2{T)
m. T > T* in each case is given by each ao(0) = line.
In agreement with the heuristic argument given earlier,
HpFhoiT) and T* are shifted to higher temperatures as
the in-plane field is directed along a gap-maximum, re-
flecting an enhanced spin depairing in this field configura-
tion. As in H II c, the FFLO state at least close to iJpFLO
has the LO-like variation. We have verified by combining
our numerical calculations with an analytical calculation
with the orbital depairing perturbatively included that
such an in-plane iJFFLO-anisotropy is absent without the
orbital depairing (i.e., when a|| = 00) and monotonously
increases with decreasing a\\ . In contrast, it is not easy to
properly predict the corresponding anisotropy (in-plane
angular dependence) of Hc2{T)-cmve: First, the depres-
sion of Hc2 due to the spin depairing is larger as the
corresponding fforb(O) is higher, and hence, the Hc2-
magnitude may not have a monotonous a||-dependence.
Second, the MF-FOT line of Hc2 is directly determined
by the details of the nonGaussian terms other than the

FIG. 2: Results corresponding to Fig.l in the cases (a) (|/3| —
0.2) and (b) (|/3| = 0.1) defined in the text.

quartic one in GL free energy and hence, is quanti-
tatively affected by our assumption of keeping the non-
Gaussian terms only up to |A|^ in eq.(l). Actually, the
rapid increase of MF-FOT line on cooling just below T*
arises due to an extremely small V6(0) near T* and might
flatten if we can numerically include the |A|* and higher
order terms. In contrast, the Vg-contribution to C2 (i.e.,
to HpFi^o{T)) was negligible, like in H || c case, consis-
tently with the smallness of Vi5(0) mentioned above. We
expect that the -ffpFLO C?") curve is less sensitive to the
neglect of the |A|^ and higher order GL terms. For these
reasons, we will focus hereafter on T* and -//fflo which
directly measure the (effective) spin depairing strength.
The resulting anisotropics of T* and -ffpFLO in Fig. 1 qual-
itatively agree with those of CeCoIns in H _L c UJ if
a gap node (or minimum) is located along [100]. As al-
ready mentioned, the MF-FOT line in = LL needs
to lie above the corresponding ai(0) = line in order for
HpFhoiT) to be realized as a transition line. As Fig.l
shows, this condition manages to be satisfied in contrast
to the H II c case.

In order to examine how the result in Fig.l is affected
by the in-plane FS anisotropy, let us next introduce it
as a Fermi velocity anisotropy in a similar manner to
Ref.El hke

VF

vf{(13) ^vf{1 + /3cos(40)),

(5)

where |/3| < 1, accompanied by the replacement A^(0)
N{0)vf/vf{(I)) in any angular integral (see eq.(4)). Ex-
cept these replacements in our calculation, the deriva-
tion of phase diagrams is quite the same as that of Fig.l.

4

When /3 > (< 0), the Fermi velocity becomes maximal
(minimal) along x. By combining these two cases with
the two candidates, y/2cos{2(j)) and •\/2sin(20), of w^, we
have four different cases of the relative anisotropics under
a fixed H || x. We will classify them into two categories,
(a) — \/2cos(2(/<) with /? < and = \/2sin(2(/))
with /3 > 0, and (b) = \/2cos(2(/)) with /3 > and
W0 = •\/2sin(2^) with^_^< 0. This classification is mo-
tivated by the result that, in the category (a), the
Fermi velocity anisotropy and the pairing anisotropy fa-
vor two different orientations, competing with each other,
of the square vortex lattices to be realized in four-fold
anisotropic d-wave superconductors in H || c, while such
a competition does not occur in (b). In Fig. 2, the re-
sulting phase diagrams for the categories (a) and (b) are
given. In the case (a), the angular dependences of iJpFLO
and T* are weakened by the FS anisotropy compared
with those in Fig. I, while the opposite tendency is seen
in the case (b). This result can be understood as fol-
lows by noting that the orbital depairing strength local
in the k-space is measured in the present case by Vy in
eq.(4) (Note that, in 2D limit, v1 is absent there). By
focusing on the case with H parallel to a gap node and
noting in the integrand of eq.(4), one will notice

that a nonzero
tions of v'j,.

tends to increase (decrease) contribu-
on average, when /? < (/3 > 0). Thus,
an enhanced orbital depairing in H parallel to a node
of the case (b) additionally reduces TJpFLO so that the
difference between the two cases in Fig. 2 follows. Bear-
ing in mind the general character of this interpretation,
we believe that the results in Fig. 2 are not qualitatively
changed by a refinement of microscopic description.

The above results commonly show a i/pFLO (2^)-liiie
shifting to higher temperatures as the in-plane field ap-
proaches a gap-maximum and, compared with the data
in CeCoIns imply a dxy-stdXe as the pairing state

of this material. Although one might consider a pos-
sibility of (ia;2_y 2 -pairing based on the fact that an ex-
tremely strong FS anisotropy in the case (a) may re-
verse the anisotropics of T* and -ffpFLOj such a strong
FS anisotropy of the case (a) should result 0| in a
square vortex lattice with an orientation due to the
FS anisotropy and hence, contradicts not only the spe-
cific heat data [Ijj but the observed orientation [13] of
H II c square vortex lattice. Therefore, an inclusion of
FS anisotropy reinforces our conclusion favoring a dxy-
pairing, although a moderate FS anisotropy competitive
in H II c with the gap anisotropy (i.e., of the case (a)) is
needed for quantitative understandings.

In conclusion, the mean field phase diagram of a type
II superconductor with strong Pauli paramagnetic de-
pairing and with a four-fold symmetric d-wave pairing
was qualitatively studied in the parallel field case. The
region in which the FFLO vortex phase appears is en-
larged when the in-plane field is directed along a gap-
maximum. This result is reinforced by including in-plane

FS anisotropics and strongly suggests a dxy pairing as the
best candidate of the gap function of CeCoIns in spite of
the electronic similarities 15j to that of high cuprates.
A reinterpretation of thermal conductivity data by Izawa
et al. Q can be seen in Ref . [l^ . The present theory
should be applicable to examining the pairing state of
other materials, such as an organic material 0, ll^ .
showing a remarkable Pauli paramagnetic depairing.

We thank Y. Matsuda, K. Machida, T. Sakakibara,
and T. Tayama for informative discussions. The present
work was finantially supported by a Grant-in- Aid from
the Ministry of Education, Culture, Sports, Science, and
Technology, Japan.

REFERENCES

[1
[2:

[4:
[6:

[9:
[10:

[11
[12:

[is:

[14

[is:

[16:
[17:

[is:

H.Shishido,
87, 057002

H. Adachi and R. Ikeda, Phys. Rev. B 68, 184510 (2003).
L.W. Gruenberg and L. Gunther, Phys. Rev. Lett. 16,
966 (1966).

M. Houzet and A. Buzdin, Phys. Rev. B 63, 184521
(2001).

K. Izawa, H.Yamaguchi, Y. Matsuda,
R.Settai, and Y.Onuki, Phys. Rev. Lett
(2001).

T. Tayama, A.Harita, T. Sakakibara, Y.Haga, H.Shishido,
R.Settai,and Y.Onuki, Phys. Rev. B 65, 180504 (2002).
A.Bianchi, R.Movshovich, N.Oeschler, P.Gegenwart,
F.Steglich, J.D.Thompson, P.G.Pagliuso, and J.L.Sarrao,
Phys. Rev. Lett. 89, 137002 (2002).

P. Fulde and R. A. Ferrell, Phys. Rev. 135, A550 (1964).
A. I. Larkin and Y. N. Ovchinnikov, Sov. Phys. JETP
20, 762 (1965).

A.Bianchi, R.Movshovich, C.Capan, P.G.Pagliuso, and
J.L.Sarrao, Phys. Rev. Lett. 91, 187004 (2003).

H. A. Radovan, N.A. Fortune,T.P.Murphy, S.T.Hannahs,
E.C.Palm, S.W.Tozer, and D.Hall, Nature 425, 51

(2003).

T.Watanabe et al., |cond-mat/0312062]

T. Tayama, private communication.

I. Kosztin and A. J. Leggett, Phys. Rev. Lett. 79, 135
(1997).

H.Aoki, T. Sakakibara, H.Shishido, R.Settai, Y.Onuki,
P.Miranovic, and K. Machida, J. Phys. Condens. Matter
16, L13 (2004).

V.A.Sidorov, M.Nicklas, P.G.Pagliuso, J.L.Sarrao,
Y.Bang, A.V.Balatsky, and J.D.Thompson, Phys. Rev.
Lett. 89, 157004 (2002).

N.Nakai, P.Miranovic, M.Ichioka, and K. Machida, Phys.
Rev. Lett. 89, 237004 (2002).

M.R. Eskildsen, C.D.Dewhurst, B.W.Hoogenboom,
C.Petrovic, and P.C.Canfield, Phys. Rev. Lett. 90,
187001 (2003).

M.A.Tanatar, T.Ishiguro, H.Tanaka, and H.Kobayashi,
Phys. Rev. B 66, 134503 (2002). Their identification of
FFLO transition line seems to be justified by a similar
result of recent thermal conductivity data [C.Capan et

5

al., 'cond-mat/0401199 in CeCoIns, if the fluctuation in [19] K.Izawa, H.Yamaguchi, T.Sasaki, and Y. Matsuda, Phys.
the organic material is strong enough to erase its discon- Rev. Lett. 88, 027002 (2002).

tinuous behavior at Hc2{T) at low T 0.